Physics 342 Lecture 6

The Infinite Square Well

Lecture 6

Physics 342
Quantum Mechanics I

Friday, February 8th, 2008

With the equation in hand, we move to simple solutions. For a particle

confined to a box, we find that the boundary conditions impose energy
quantization (specific allowed energies), a new phenomenon with respect to
classical mechanics in a box. This solution comes to us in a familiar way via
separation of variables (think of separable solutions to Laplace’s equation
representing the electrostatic potential inside a box with grounded walls).

6.1 Separability of Schrödinger’s Equation

For a time-independent potential, V (x), Schrödinger’s equation is separable

in time, i.e.
∂Ψ ~2 ∂ 2 Ψ
i~ =− + V (x) Ψ (6.1)
∂t 2 m ∂x2
with the ansatz: Ψ(x, t) = ψ(x) φ(t) separates (notice that for V (x, t) this
may or may not be true)

φ̇ ~2 ψ 00
i~ =− + V (x) (6.2)
φ 2m ψ

where we define φ̇(t) ≡ dφ(t)

dt and ψ 0 (x) ≡ dψ(x)
dx . Then the right-hand-
side depends only on x, the left-hand side depends only on t, so they must
separately be equal to a constant, call it E. The time-dependence of the
wave function is then completely fixed:
E
i ~ φ̇ = E φ −→ φ = A e−i ~
t
. (6.3)

1 of 12
6.1. SEPARABILITY OF SCHRÖDINGER’S EQUATION Lecture 6

We then have the so-called time-independent Schrödinger equation, which

is just the spatial portion:
~2 ψ 00
− + V (x) = E. (6.4)
2m ψ

Notice that our full solution, whatever the solution to (6.4) is, has time-
independent density:
|Ψ(x, t)|2 = Ψ(x, t)∗ Ψ(x, t) = ψ ∗ (x) ψ(x), (6.5)
R∞
and the probabilistic normalization −∞ |Ψ(x, t)|2 dx = 1 becomes an issue
for the spatial part alone.
As we saw at the end of last time, the spatial component of Schrödinger’s
equation can be viewed as the operator constructed from a Hamiltonian –
2
that is, H(x, p) = 2pm + V (x) becomes

~ ∂ ~2 ∂ 2
 
Ĥ x, =− + V (x) (6.6)
i ∂x 2 m ∂x2

and we can write the separated form as:

Ĥ ψ(x) = E ψ(x). (6.7)
This looks an awful lot like an eigenvector equation for the operator Ĥ with
eigenfunction ψ(x) and eigenvalue E (that’s what it is).
We can calculate the expectation value of Ĥ from the above, assuming the
wavefunction has been normalized:
Z ∞ Z ∞
∗
hĤi = ψ (x) Ĥ ψ(x) dx = E ψ ∗ (x) ψ(x) dx = E, (6.8)
−∞ −∞

so we say that the state represented by ψ(x) has energy E. This is clearly
a constant in time – what’s more surprising, the variance of the energy is
zero – i.e. the particle in state ψ(x) has exactly energy E:
σ 2 = hĤ 2 i − hĤi2 = E 2 − E 2 = 0. (6.9)

Currently, E is continuous, we have no restriction on what its value can

be – but in many cases, the energy itself is quantized (we will see this in a
moment), so that there are a variety of states ψE (x) with definite energy, and
any linear combination of these is also a solution to the time-independent
Schrödinger equation.

2 of 12
6.2. INFINITE SQUARE WELL Lecture 6

6.2 Infinite Square Well

Consider a particle confined to a box in one dimension. In this case, the

confining potential takes the form:
0 0≤x≤a

V (x) = (6.10)
∞ x < 0 or x > a

In the region 0 ≤ x ≤ a, our eigenfunction equation: Ĥ ψ = E ψ reads:

√ √
−~2 00 2mE 2mE
ψ (x) = E ψ(x) −→ ψ(x) = A ei ~ + B e−i ~ . (6.11)
2m
We have boundary conditions – we cannot have any ψ to the left of x = 0
or to the right of x = a. This first condition gives:

ψ(x = 0) = A + B = 0 −→ A = −B, (6.12)

at which point we can write the solution in the form:

√ !
2mE
ψ(x) = Ã sin x , (6.13)
~

for some new (but still arbitrary) constant à (= 2 i A). Then the boundary
condition at x = a gives
√ !
2mE
ψ(x = a) = 0 = Ã sin a (6.14)
~

with trivial solution à = 0, which gives ψ(x) = 0 and is not normalizable.

The more interesting (viable) case is:
√
2mE
a = nπ for n ∈ Z. (6.15)
~
The energy of the particle, then, can take only particular values, in integer
multiples:
~2 n2 π 2
En = . (6.16)
2 m a2
So define the infinite set of solutions:
2 n π x
r
ψn (x) = sin , (6.17)
a a

3 of 12
6.2. INFINITE SQUARE WELL Lecture 6

where the normalization factor is for later convenience.

We see that the energy for the state ψn (x) is directly related to its wave-
length. Since we have a finite domain, that wavelength defines the number
of cycles that can fit in our interval, and the energy is a measure of the
number of full cycles as shown in Figure 6.1.

n=4

n=3
Energy

n=2

n=1

Figure 6.1: The first four infinite square well energy eigenfunctions. The
number of extrema is related to the energy (∼ n2 )

The most general spatial solution is a sum:

∞ ∞
2 n π x
r
X X
ψ(x) = cn ψn (x) = cn sin . (6.18)
a a
n=1 n=1

Now, as we discussed in the linear algebra section, we can view ψn (x) as a

basis for functions that are periodic – the inner product that is relevant is:
a 2
Z a  
ψn · ψm = ψn (x) ψm (x) dx = δmn = δmn (6.19)
0 2 a
q
(that naked δmn is the motivation for the funny factor of a2 in (6.17)). If
we were given an initial wavefunction ψ0 (x), we could define the coefficients

4 of 12
6.2. INFINITE SQUARE WELL Lecture 6

cn in the obvious way:

Z a
cn = ψ0 (x) ψn (x)dx for n = 1 −→ ∞. (6.20)
0

In
R a addition, we know that such an initial waveform must be normalized:
0 0ψ (x)2 dx = 1. Suppose we had the coefficients c , then:
n

∞
!2 ∞
Z a X X
cn ψn (x) dx = c2n = 1 (6.21)
0 n=1 n=1

is the requirement.
The full wave-function is built up out of the infinite spatial expansion with
appropriate temporal evolution term:
∞ ∞
2 n π x
r
2 2
−i E~n t −i ~ n π 2 t
X X
Ψ(x, t) = cn e ψn (x) = cn e 2 m a sin , (6.22)
a a
n=1 n=1

6.2.1 Examples

The Schrödinger equation is first order in time, hence the need to specify an
initial waveform at t = 0. That gives us the spatial decomposition discussed
above, and familiar from almost all treatments of separation of variables
(where the solutions form a “complete” basis). Suppose we start with a
pure n = 1 solution, that is, we require that Ψ(x, 0) = A sin πax – here we

don’t need any fancy integration, the solution is obvious:

~ π2 t
π x
Ψ(x, t) = A e−i 2 m a2 sin . (6.23)
a
R∞
Now, we require that −∞ |Ψ(x, t)|2 dx = 1, and that can be used to set the
constant A:

2 πx 2 a 2
Z ∞ Z a   r
∗
Ψ Ψ dx = A sin
2
dx = A = 1 −→ A = , (6.24)
−∞ 0 a 2 a

and we have the appropriately normalized final form

2 −i ~ π2 t2 π x
r
Ψ(x, t) = e 2 m a sin . (6.25)
a a

5 of 12
6.2. INFINITE SQUARE WELL Lecture 6

From here, we can calculate the expectation value of three of our operators
of interest: x, p and H, giving us the average position, momentum and
energy of the particle:

a
Z ∞
hxi = Ψ∗ (x, t) x Ψ(x, t) dx =
−∞ 2
~ ∂
Z ∞  
hpi = Ψ∗ (x, t) Ψ(x, t)dx = 0 (6.26)
−∞ i ∂x
~2 π 2
Z ∞
~2
 
∗
hHi = Ψ (x, t) − Ψ00 (x, t)dx = .
−∞ 2m 2 m a2

The expectation value of position is at half the width of the confinement,

which is sensible. The energy expectation value is also precisely what we
expect.
If we think of the “classical” version of this problem, a particle moving back
and forth with constant momentum, the expectation value of the momen-
tum would be zero, in the sense that the particle spends equal time moving
to the left and the right. To find the (classical) expectation value of po-
sition, we must know the time-independent position density (the analogue
of |Ψ(x, t)|2 ). The particle moves back and forth, so technically, at a given
time, its position is known and the density is a delta function about the
actual location of the particle: ρ(x, t) = δ(x − x(t)). In this case, we want
the pure spatial density, so we average over “one full cycle” in time. Con-
sider the trip from x = 0 to x = a at constant velocity. For this segment of
the motion, we go from t = 0 → T ≡ a/v with x(t) = v t. As we go from
x = a → 0, we have t = T → 2 T with x(t) = −a + v (t − T ). Then if we
average over a round trip:

1
Z T Z 2T 
ρ(x) = ρ(x, t) dt + ρ(x, t) dt . (6.27)
2T 0 T

We just need to input and evaluate the delta functions for the two different
trajectories, then

1
Z T Z 2T 
ρ(x) = δ(x − v t) dt + δ(x − a + v (t − T )) dt
2T 0 T
1 1 1
Z x−a   Z x   
= δ(y) − dy + δ(y) dy (6.28)
2T x v x−a v
1 1
=
T v

6 of 12
6.2. INFINITE SQUARE WELL Lecture 6

where the second equality follows from a change of variables in each inte-
gration. Using the definition of the travel time: v T = a, we learn that
ρ(x) = a1 , i.e. the probability that the particle is in an interval dx is propor-
tional to dx (ρ dx = dx/a) – that makes sense, the larger the interval, the
more likelyRthe particle isR in it. In addition, the normalization has come out
a a
correctly: 0 ρ(x) dx = 0 a1 dx = 1, so we are bound to find the particle
somewhere between the walls.
From this classical density, it is clear that the expectation value of position
is hxi = a2 , identical to the quantum prediction1 . So the only real difference
between the expected values associated with a stationary solution ψn (x) of
Schrödinger’s equation and the classical expectation values is the restriction
to a finite set of energies – this does not hold on the classical side.
There are other differences – suppose we look at the variance for the position,
momentum and energy operators for n = 1. For the quantum mechanical
density, these are (remember that the variance of an operator Q is σQ 2 =

hQ2 i − hQi2 )
a2
σx2 = (π 2 − 6)
12 π 2
~2 π 2 (6.29)
σp2 =
a2

σE2
= 0.
The last relation tells us that we are in a state of definite energy (no width
to the energy expectation value) – an energy eigenstate.
For the classical quantities, we have constant energy, so the variance will be
zero. The classical position variance can be found using the classical density,
2
it ends up being σx2 = a12 , the first term of the quantum form. If we were to
calculate the variances using a generic eigenstate, labelled by n, we would
find:
a2 a2
σx2 = −
12 2 n2 π 2
n2 ~2 π 2 (6.30)
σp2 =
a2
σE2
= 0.
These energy eigenstates all have definite energy, and from the variance for
position, we can see that in the limit n −→ ∞, we recover the classical
1 1
We calculated this quantum hxi = 2
a for n = 1, but this is the position expectation
value for all n.

7 of 12
6.2. INFINITE SQUARE WELL Lecture 6

distribution.

6.2.2 Mixed States

The above discussion is highly specialized to eigenstates of the Hamiltonian

– as we have seen, these have time-independent expectation values and well-
defined energies (although only a “small” set are allowed). Suppose we
prepare (in some manner) a system that is an equal admixture of the n and
` states:  E` t 
En t
Ψ(x, t) = cn e−i ~ ψn (x) + c` e−i ~ ψ` (x) . (6.31)

We must first choose cn and c` to normalize the wavefunction:

Z a
En t E` t 
Ψ(x, t) · Ψ(x, t) = c∗n ei ~ ψn (x)∗ + c∗` ei ~ ψ` (x)∗
0 En t E` t 
× cn e−i ~ ψn (x) + c` e−i ~ ψ` (x) dx

(En −E` ) t
= c∗n cn ψn (x) · ψn (x) + c∗n c` ei ~ ψn (x) · ψ` (x)

(E −E ) t
∗ i ` ~ n ∗ ∗
+ c` e ψ` (x) · ψn (x) + c` c` ψ` (x) · ψ` (x) ,
(6.32)
and keeping in mind that ψn (x) · ψ` (x) = δn` , we can simplify:

Ψ(x, t) · Ψ(x, t) = c∗n cn + c∗` c` . (6.33)

Since we have already decided to make our state an equal admixture, we

must have cn = c` = √12 . Our final form is

1  En t E` t 
Ψ(x, t) = √ e−i ~ ψn (x) + e−i ~ ψ` (x) . (6.34)
2

To calculate the expectation value for energy, we have

1 a ∗
Z
hHi = Ψ (x, t) Ĥ Ψ(x, t) dx
2 0
(6.35)
1 a ∗
Z  En t E` t 
= Ψ (x, t) En e−i ~ ψn (x) + E` e−i ~ ψ` (x) dx.
a 0
We have used Schrödinger’s equation to simplify the action of the operator
H on Ψ(x, t) – since the wave function is already decomposed in terms of

8 of 12
6.2. INFINITE SQUARE WELL Lecture 6

eigenstates. In dot-product notation, the above is

1
  (E −E ) (E −E ) 
i n~ ` −i n ~ `
hHi = En ψn (x) · ψn (x) + E` e + En e ψn (x) · ψ` (x)
2

+ E` ψ` (x) · ψ` (x)
1
= (En + E` ).
2
(6.36)
It is possible to pick an eigenstate with the same expectation value for the
energy, and it is interesting to ask how one might distinguish between such
a state and the above. For example, if we set n = 7, ` = 1, then the
expectation value for E is:
~2 π 2 50
, (6.37)
2 m a2 2
and this is identical to hHi for n = 5.
It is the variance that makes the difference – for a pure n = 5 state, the
variance of the energy is zero, it is an eigenstate of the Hamiltonian. In the
mixed state setting, we have:

2
1
1 2 ~4 π 4 `2 − n2
σH =
2
E + En − (E` + En ) =
2 2
, (6.38)
2 ` 4 16 a4 m4
and we see that these states do not have definite energy – there is a spread,
depending on the eigenstates making up Ψ(x, t).

Example

To see how the mixed states lead to time-evolution, we can pick an

arbitrary (but normalized) initial state, decompose it into the sine basis
that are the eigenvectors of the Hamiltonian operator for the infinite
square well, and form Ψ(x, t) explicitly. Start with
2
Ψ(x, 0) = A eβ (x−d) x (x − a), (6.39)

so that the initial wavefunction

R d satisfies the boundary conditions, and
we can use A to ensure that 0 Ψ(x, 0) Ψ(x, 0) dx = 1. This initial dis-
∗

tribution represents a Gaussian peaked about the value d, and vanishing

at x = 0, a, the walls of our box.

9 of 12
6.3. MEASUREMENT Lecture 6

We calculate the coefficients in the expansion of the wavefunction in the

usual way from (6.20). Then all we have to do is input these coeffi-
cients in (6.22). We can plot the resulting ρ(x, t) = Ψ∗ (x, t) Ψ(x, t) as a
function of time, and a few snapshots are shown below:

25 25 25

20 20 20

15 15 15

10 10 10

5 5 5

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

25 25 25

20 20 20

15 15 15
Out[14]=
10 10 10

5 5 5

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

25 25 25

20 20 20

15 15 15

10 10 10

5 5 5

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

6.3 Measurement

The final input into our quantum mechanical picture is a statement of what a
measurement does. What a measurement is is another question – if you think
about it, though, most measurements involve interaction. Those interactions
can be carried out using other particles (so we have multiple particles) or
turning on and off various detection equipment that measures properties
of a particle (effectively introducing time-dependent potential interactions).
These considerations take us away from our static V (x) potentials acting on
single particles, and so are difficult to analyze at this stage.
What a measurement does is defined with reproducability in mind. For
an operator like Ĥ, corresponding to the observable “energy”, we say that
a measurement of energy puts the system into an energy eigenstate and
returns one of the energy eigenvalues. Therein lies part of the “magic” of
quantum mechanics. Never mind how this happens. Think about the dull
implications for our particle in a box. When we measure the energy of a
generic state like (6.22), we get a measured value that must be one of the
eigenvalues:
n2 π 2 ~2
E= (6.40)
2 m a2

10 of 12
6.3. MEASUREMENT Lecture 6

and we subsequently know that the particle is in the nth eigenstate of Ĥ. Of
course, once the particle is an an eigenstate of Ĥ, its energy does not change.
So an immediately repeated measurement will return the same energy value
and leave the particle in the same state.
Given a general solution like (6.22), we can compute the expectation value
of energy in terms of the decomposition coefficients (exploiting the inner
product relation for the ψm (x) from (6.19)):
∞
!∗
2 n π x
Z ∞ X r
En t
hHi = cn sin e−i ~
−∞ n=1 a a
∞
!
2 m π x
r
−i Em t
(6.41)
X
× Em cm sin e ~
a a
m=1
∞
X
= Em |cm |2 .
m=1

This is reminiscent of the discrete probability expectation values we dis-

cussed, in which we had averages like: hf (j)i = ∞ f (j) P (j), and sug-
P
j=0
gests that we might interpret the coefficients |cm | as probabilities. This
2

idea is supported by the normalization of the wavefunction – from (6.21),

we must have:
X∞
|cm |2 = 1, (6.42)
m=1

which is also required if we want to understand |cm |2 as P (m). For a generic

state like this, we say that a measurement of energy will return the value
En with probability |cn |2 and upon measurement, the wavefunction is in
the stationary state Ψn (x, t). This interpretation is also reasonable from a
linear algebra point of view – the decomposition coefficients cm tell us how
much Ψm (x, t) is in Ψ(x, t), and so it is easy to predict the state selected by
measurement.
The story is only mildly different for other operators – if we make a position
measurement, we will recover one of the eigenvalues of the position opera-
tor2 , these do evolve in time (as it turns out), but an immediately repeated
measurement of position must return the same location.
2
Position, as an operator, has a continuous spectrum, and its eigenvectors are Dirac
delta functions, as we shall see later on.

11 of 12
6.3. MEASUREMENT Lecture 6

Homework

Reading: Griffiths, pp. 24–38.

Problem 6.1

Knock ’em dead.

12 of 12